Light transport algorithms based on ray tracing are used to sample a three dimensional scene and its illumination by casting rays within the scene. Typically, a large number of rays are required for high quality rendering, i.e. with an acceptable noise level. Because of this complexity, it is often beneficial to perform parts of lighting computations using a simplified representation of the three dimensional scene. Typical simplified representations are created using a standard k-nearest neighbor search to adapt to the local density of ray endpoints and derive a radius for each one. However, while creating a simplified representation is fast and automatic, there is no guarantee that the obtained simplified representation faithfully represents the original three dimensional scene. Scene misrepresentations due to simplification can lead to unwanted artifacts in rendered images.